Post-instrument covariates are often included as controls in instrumental variable (IV) analyses to address a violation of the exclusion restriction. However, we show that such analyses are subject to biases unless strong assumptions hold. Using linear constant-effects models, we present asymptotic bias formulas for three estimators (with and…

When estimating causal effects, unmeasured confounding and model misspecification are both potential sources of bias. We propose a method to simultaneously address both issues in the form of a semi-parametric sensitivity analysis. In particular, our approach incorporates Bayesian Additive Regression Trees into a two-parameter sensitivity analysis…

Propensity score analysis (PSA) is a methodological technique which may correct for selection bias in a quasi-experiment by modeling the selection process using observed covariates. Because logistic regression is well understood by researchers in a variety of fields and easy to implement in a number of popular software packages, it has…

Causal models often omit variables that should be included, use variables that are measured fallibly, and ignore time lags. Such practices can lead to severely biased estimates of effects. The discussion explains these biases and shows how to take them into account. (Author)

There are difficulties in expressing effect sizes on a common metric when some studies use transformed scales to express group differences, or use factorial designs or covariance adjustments to obtain a reduced error term. A common metric on which effect sizes may be standardized is described. (Author/RL)

In large-scale surveys, it is no longer uncommon for repeated measurements to be obtained from respondents and analyses performed to gauge the magnitiude of nonsampling errors. This is particularly true for periodic surveys and longitudinal surveys where a very large investment in data collection is made. This technical note, aimed at the analysis…

Single administration (beta-binomial) estimates for the raw agreement index p and the corrected-for-chance kappa index in mastery testing are compared with those based on two test administrations in terms of estimation bias and sampling variability. Bias is about 2.5 percent for p and 10 percent for kappa. (Author/RL)

W. M. K. Trochim and others defend the record of the regression-discontinuity (RD) design and blur the statistical tests for treatment effect. Their Monte Carlo results show the problematic nature of RD and its potential bias. New testing strategies and restrictions for the application of RD are proposed. (SLD)

A logistic model was used to generate data to serve as a proxy for an immediate retest from item responses to a fourth grade standardized reading comprehension test of 45 items. Assuming that the actual test may be considered a pretest and the proxy data may be considered a retest, the effect of regression was investigated using a percentage of…

The results of six major projects are discussed including a comprehensive mathematical and statistical analysis of the problems caused by errors of measurement in linear models for assessing change. In a general matrix representation of the problem, several new analytic results are proved concerning the parameters which affect bias in…

The regression-discontinuity design involving a treatment interaction effect (TIE), pretest-posttest functional form specification, and choice of point-of-estimation of the TIE are examined. Formulas for controlling the magnitude of TIE in simulations can be used for simulating the randomized experimental case where estimation is not at the…

The effects of the violation of the assumption of normality coupled with the condition of multicollinearity upon the outcome of testing the hypothesis Beta equals zero in the two-predictor regression equation is investigated. A monte carlo approach was utilized in which three differenct distributions were sampled for two sample sizes over…

In an experiment manipulating artificial data in a factorial design, model misspecification and varying levels of error in measurement and in model structure are shown to have significant effects on LISREL parameter estimates in a modified peer influence model. (Author/LMO)

How accurately ability is estimated when the test model does not fit the data is considered. To address this question, this study investigated the accuracy of the maximum likelihood estimator of ability for the one-, two- and three-parameter logistic (PL) models. The models were fitted into generated item characteristic curves derived from the…

The RMC Research Corporation evaluation model C1--the special regression model (SRM)--was evaluated through a series of computer simulations and compared with an alternative model, the norm referenced model (NRM). Using local data and national norm data to determine reasonable values for sample size and pretest posttest correlation parameters, the…